bpf: Add support for forced LINK_DETACH command

Add LINK_DETACH command to force-detach bpf_link without destroying it. It has
the same behavior as auto-detaching of bpf_link due to cgroup dying for
bpf_cgroup_link or net_device being destroyed for bpf_xdp_link. In such case,
bpf_link is still a valid kernel object, but is defuncts and doesn't hold BPF
program attached to corresponding BPF hook. This functionality allows users
with enough access rights to manually force-detach attached bpf_link without
killing respective owner process.

This patch implements LINK_DETACH for cgroup, xdp, and netns links, mostly
re-using existing link release handling code.

Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20200731182830.286260-2-andriin@fb.com

include/linux/bpf.h

@@ -793,6 +793,7 @@ struct bpf_link {
 struct bpf_link_ops {
 	void (*release)(struct bpf_link *link);
 	void (*dealloc)(struct bpf_link *link);
+	int (*detach)(struct bpf_link *link);
 	int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
 			   struct bpf_prog *old_prog);
 	void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);

include/uapi/linux/bpf.h

@@ -117,6 +117,7 @@ enum bpf_cmd {
 	BPF_LINK_GET_NEXT_ID,
 	BPF_ENABLE_STATS,
 	BPF_ITER_CREATE,
+	BPF_LINK_DETACH,
 };
 
 enum bpf_map_type {
@@ -634,6 +635,10 @@ union bpf_attr {
 		__u32		old_prog_fd;
 	} link_update;
 
+	struct {
+		__u32		link_fd;
+	} link_detach;
+
 	struct { /* struct used by BPF_ENABLE_STATS command */
 		__u32		type;
 	} enable_stats;

kernel/bpf/cgroup.c

@@ -814,6 +814,7 @@ static void bpf_cgroup_link_release(struct bpf_link *link)
 {
 	struct bpf_cgroup_link *cg_link =
 		container_of(link, struct bpf_cgroup_link, link);
+	struct cgroup *cg;
 
 	/* link might have been auto-detached by dying cgroup already,
 	 * in that case our work is done here
@@ -832,8 +833,12 @@ static void bpf_cgroup_link_release(struct bpf_link *link)
 	WARN_ON(__cgroup_bpf_detach(cg_link->cgroup, NULL, cg_link,
 				    cg_link->type));
 
+	cg = cg_link->cgroup;
+	cg_link->cgroup = NULL;
+
 	mutex_unlock(&cgroup_mutex);
-	cgroup_put(cg_link->cgroup);
+
+	cgroup_put(cg);
 }
 
 static void bpf_cgroup_link_dealloc(struct bpf_link *link)
@@ -844,6 +849,13 @@ static void bpf_cgroup_link_dealloc(struct bpf_link *link)
 	kfree(cg_link);
 }
 
+static int bpf_cgroup_link_detach(struct bpf_link *link)
+{
+	bpf_cgroup_link_release(link);
+
+	return 0;
+}
+
 static void bpf_cgroup_link_show_fdinfo(const struct bpf_link *link,
 					struct seq_file *seq)
 {
@@ -883,6 +895,7 @@ static int bpf_cgroup_link_fill_link_info(const struct bpf_link *link,
 static const struct bpf_link_ops bpf_cgroup_link_lops = {
 	.release = bpf_cgroup_link_release,
 	.dealloc = bpf_cgroup_link_dealloc,
+	.detach = bpf_cgroup_link_detach,
 	.update_prog = cgroup_bpf_replace,
 	.show_fdinfo = bpf_cgroup_link_show_fdinfo,
 	.fill_link_info = bpf_cgroup_link_fill_link_info,

kernel/bpf/net_namespace.c

@@ -142,9 +142,16 @@ static void bpf_netns_link_release(struct bpf_link *link)
 	bpf_prog_array_free(old_array);
 
 out_unlock:
+	net_link->net = NULL;
 	mutex_unlock(&netns_bpf_mutex);
 }
 
+static int bpf_netns_link_detach(struct bpf_link *link)
+{
+	bpf_netns_link_release(link);
+	return 0;
+}
+
 static void bpf_netns_link_dealloc(struct bpf_link *link)
 {
 	struct bpf_netns_link *net_link =
@@ -228,6 +235,7 @@ static void bpf_netns_link_show_fdinfo(const struct bpf_link *link,
 static const struct bpf_link_ops bpf_netns_link_ops = {
 	.release = bpf_netns_link_release,
 	.dealloc = bpf_netns_link_dealloc,
+	.detach = bpf_netns_link_detach,
 	.update_prog = bpf_netns_link_update_prog,
 	.fill_link_info = bpf_netns_link_fill_info,
 	.show_fdinfo = bpf_netns_link_show_fdinfo,

kernel/bpf/syscall.c

@@ -3991,6 +3991,29 @@ out_put_link:
 	return ret;
 }
 
+#define BPF_LINK_DETACH_LAST_FIELD link_detach.link_fd
+
+static int link_detach(union bpf_attr *attr)
+{
+	struct bpf_link *link;
+	int ret;
+
+	if (CHECK_ATTR(BPF_LINK_DETACH))
+		return -EINVAL;
+
+	link = bpf_link_get_from_fd(attr->link_detach.link_fd);
+	if (IS_ERR(link))
+		return PTR_ERR(link);
+
+	if (link->ops->detach)
+		ret = link->ops->detach(link);
+	else
+		ret = -EOPNOTSUPP;
+
+	bpf_link_put(link);
+	return ret;
+}
+
 static int bpf_link_inc_not_zero(struct bpf_link *link)
 {
 	return atomic64_fetch_add_unless(&link->refcnt, 1, 0) ? 0 : -ENOENT;
@@ -4240,6 +4263,9 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
 	case BPF_ITER_CREATE:
 		err = bpf_iter_create(&attr);
 		break;
+	case BPF_LINK_DETACH:
+		err = link_detach(&attr);
+		break;
 	default:
 		err = -EINVAL;
 		break;

net/core/dev.c

@@ -8979,12 +8979,20 @@ static void bpf_xdp_link_release(struct bpf_link *link)
 	/* if racing with net_device's tear down, xdp_link->dev might be
 	 * already NULL, in which case link was already auto-detached
 	 */
-	if (xdp_link->dev)
+	if (xdp_link->dev) {
 		WARN_ON(dev_xdp_detach_link(xdp_link->dev, NULL, xdp_link));
+		xdp_link->dev = NULL;
+	}
 
 	rtnl_unlock();
 }
 
+static int bpf_xdp_link_detach(struct bpf_link *link)
+{
+	bpf_xdp_link_release(link);
+	return 0;
+}
+
 static void bpf_xdp_link_dealloc(struct bpf_link *link)
 {
 	struct bpf_xdp_link *xdp_link = container_of(link, struct bpf_xdp_link, link);
@@ -9066,6 +9074,7 @@ out_unlock:
 static const struct bpf_link_ops bpf_xdp_link_lops = {
 	.release = bpf_xdp_link_release,
 	.dealloc = bpf_xdp_link_dealloc,
+	.detach = bpf_xdp_link_detach,
 	.show_fdinfo = bpf_xdp_link_show_fdinfo,
 	.fill_link_info = bpf_xdp_link_fill_link_info,
 	.update_prog = bpf_xdp_link_update,